Skip to main content

What Is Research & Explore?

Research & Explore is a conversational interface to your archive. You pick a scope (one accession, several, a dossier, or your whole archive), ask questions in plain language, and get cited answers grounded in your material.

Available on Starter and above.

A new Explore session: the accessions in scope, a sentence saying how many items the conversation can see, and starter questions


SearchExplore
Best forFinding itemsFinding answers
InputKeywords, filtersQuestions in natural language
OutputResult cards with metadataA composed answer with citations and passages
StrengthsExact matches, faceted browsingSynthesis across many items, follow-ups, verbatim evidence
LimitationsDoesn't composeSlower; can sometimes mis-cite

Search and Explore complement each other. A typical flow: search to find candidate items, then start an Explore session on those items to ask follow-up questions.

Each answer in Explore grounds itself in specific passages from your material — marked excerpts with locators (page numbers for documents, timestamps for audio and video). This means you can see exactly where the answer comes from and jump straight to the source.


What Explore can do

  • Answer questions that span several items — "What does the correspondence say about the 1956 dispute?"
  • Build timelines from heterogeneous sources — "List every meeting where the Smith case was raised, in chronological order."
  • Surface entities — "Who else is named alongside Mary Thornton in these letters?"
  • Find gaps — "Are there years between 1880 and 1920 where I have no material from this family?"
  • Compose narrative — "Summarise the development of housing policy in this fond"

Each answer cites the items it drew from; you can click through to see the source material.

Opening the Sources panel under an answer to reveal the eight records the answer drew on, with the "The AI wrote this. Check it before you rely on it." marker below


What Explore can't do

  • Answer questions about anything outside your archive. It knows what you've uploaded; it doesn't know the rest of history.
  • Read un-OCR'd material. If a document hasn't been transcribed, its content is invisible to the model.
  • Be 100% accurate. The model can mis-cite, miss nuance, or over-generalise. Treat answers as research leads, not final claims.

Languages and access

Questions in any language. Ask in English, Welsh, French, German, or any other language in which you'd normally research. Explore will answer in the same language the question was asked.

Workspace-wide control. A workspace admin can switch Explore and dossier analysis off for the whole workspace under Settings → Organisation. With it off, the Explore page says so and offers nothing to run. Explore is also for archivists and admins — reviewers and volunteers are told as much rather than shown a locked screen.


Plan limits

PlanSessions / monthMessages / sessionAccessions / session
Community0 (not available)
Starter2050500
Professional50100500
InstitutionUnlimited200500

Each conversation is one session. Each message you send counts toward the per-session message cap. Starting a conversation is free; each question you ask uses 1 credit.


What's next?